A number of studies have suggested that type of dialysis membrane is associated with differences in long-term outcome of patients undergoing hemodialysis, both in terms of morbidity and mortality. The purpose of this study was to determine the relationship of membrane type and specific causes of death. Data from the United States Renal Data System Case Mix Adequacy Study, a national random sample of hemodialysis patients who were alive on December 31, 1990, were used. Our study was limited to patients in this data set who were undergoing dialysis for at least 1 year (n = 4,055). For the main analytic models, membrane type was classified into two categories: unmodified cellulose or MC/SYN (which combines modified cellulose [MC] and synthetic membranes [SYN]). The relationships of membrane type and major causes of mortality were analyzed using Cox proportional hazards models, which adjusted for multiple (21) covariates, including demographics, comorbidity, Kt/V, and other parameters. Patients were censored at transplantation or 60 days after a switch to peritoneal dialysis. Compared with patients dialyzed with unmodified cellulose membranes, the adjusted relative mortality risk (RR) from infection was 31% lower (RR = 0.69; P = 0.03) and from coronary artery disease was 26% lower (RR = 0.74; P = 0.07) for patients dialyzed with MC/SYN membranes. No statistically significant difference (all P > 0.1) was found in mortality risk from cerebrovascular disease (RR = 1.08), other cardiac causes (RR = 0.86), malignancy (RR = 0.90), or other known causes (RR = 0.82) between patients dialyzed with MC/SYN compared with unmodified cellulose membranes. These results offer support to reported experimental and observational clinical studies that have found that unmodified cellulose membranes may increase the risk for both infection and atherogenesis. Further studies are necessary to evaluate the possibility of confounding factors, compare more specific membrane types, and determine the pathophysiology linking membrane type to cause-specific mortality.